Mini-batch Coresets for Memory-efficient Language Model Training on Data Mixtures.

Dang Nguyen, Wenhan Yang, Rathul Anand, Yu Yang, Baharan Mirzasoleiman

Browse the full ICLR paper archive.